Prepare input file for Diels-Alder reaction

DeepCV's input file must be in a JSON file format (dictionary-like). An example of DAENN input in inputs/ folder shows the configuration for training a model using DAENN with five hidden layers. The first three hidden layers contain two encoded layers and one latent encoded layer (middle layer). The rest layers are two decoded layers for reconstruction. On the other hand, the size of two hidden layers that are opposite of each other, e.g., input and output layers (the 1^st and 5^th hidden layers) must be the same.

Keys

Project

Key	Definition	Value
`name`	Project name	String
`neural_network`	Type of neural network	`daenn`, `gan`

Dataset

Key	Definition	Value
`primary`	A list of dataset files for primary loss function	String
`secondary`	A list of dataset files for secondary loss function	String
`split`	Split dataset	Logical: `true`, `false`
`split_ratio`	Ratio for splitting (for training set)	Integer: `0.8`
`shuffle`	Shuffle the data points	Logical: `true`, `false`
`normalize_scale`	Normalization scaling value	Float: `0.0`
`max_scale`	Maximum scaling value	Integer: `1`

Model

Key	Definition	Value
`optimizer`	Optimizer	`Adadelta`, `Adagrad`, `Adam`, `Adamax`, `Ftrl`, `Nadam`, `RMSprop`, `SGD`
`main_loss`	Primary loss function	`MSE`, `MAE`
`penalty_loss`	Secondary loss function	`MSE`, `MAE`
`loss_weights`	A list of weight for each loss	`[1, 0.1]`
`num_epoch`	Number of training epoch	Integer: `1000`
`batch_size`	Batch size	Integer: `55`

Neural network

Key	Definition	Value
`units`	A list of number of neurons per hidden layer	Integer
`act_funcs`	A list of activation function for each hidden layer	`relu`, `sigmoid`, `tanh`

Performance

Key	Definition	Value
`enable_gpu`	Train on GPU	Logical: `true`, `false`
`gpus`	Number of GPUs	Integer

Settings

Key	Definition	Value
`verbosity`	Level of output printing	Integer: `1`
`show_summary`	Show DAENN summary	Logical: `true`, `false`
`save_tb`	Save TensorBoard file	Logical: `true`, `false`
`save_model`	Save trained model	Logical: `true`, `false`
`save_weights`	Save weights	Logical: `true`, `false`
`save_weights_npz`	Save weights in npz format	Logical: `true`, `false`
`save_biases_npz`	Save biases in npz format	Logical: `true`, `false`
`save_graph`	Save TensorFlow graph	Logical: `true`, `false`
`save_loss`	Save loss	Logical: `true`, `false`
`show_loss`	Show loss	Logical: `true`, `false`
`save_metrics`	Save metrics	Logical: `true`, `false`
`show_metrics`	Show metrics	Logical: `true`, `false`

Output

Key	Definition	Value
`out_dir`	Path for output directory	String
`out_model`	Name of output model	String
`out_weights`	Name of output weights	String
`out_weights_npz`	Name of output weights in npz format	String
`out_biases_npz`	Name of output biases in npz format	String
`loss_plot`	Name of loss plot	String
`metrics_plot`	Name of metrics plot	String